AI for English speaking practice: A study of effectiveness and engagement among Vietnamese university learners

This study investigated the effectiveness of artificial intelligence- (AI) powered speaking practice tools in improving English proficiency and engagement among Vietnamese university students using a 16-week quasi-experimental design with 240 participants across four diverse higher education institutions. Multi-level modelling revealed significantly greater speaking proficiency improvements for the treatment group compared to the control group (overall d = 1.11), with substantial gains in learner engagement across all dimensions. Although the intervention proved effective across all institutional types, effect sizes varied between well-resourced and moderately resourced rural institutions, suggesting resources moderate rather than determine outcomes. The findings demonstrate AI tools' potential to address persistent challenges in Vietnamese English education by creating low-stakes practice environments that circumvent cultural barriers to speaking participation. The study extends theoretical understanding by establishing that AI tools primarily enhance performance-oriented aspects of speaking over knowledge-based dimensions and by identifying cognitive engagement as a critical mechanism underlying effective language acquisition in Confucian-heritage educational contexts. Implications for practice or policy: Language educators should integrate AI-powered speaking tools to provide personalised practice opportunities, addressing traditional classroom resource constraints. Vietnamese higher education institutions must ensure equitable AI tool access across socio-economic backgrounds while maintaining cultural pedagogical appropriateness. Curriculum designers should revise speaking assessments to capture both test performance and authentic communicative competence improvements. Policymakers should prioritise teacher professional development for effective AI integration in language instruction.